The 100k World Record is held by Nao Kazami at 6:09:14. When he ran this in 2018 his personal best at the marathon was a mere 2:17:23, but he has since run 2:13:13 (Tokyo 2020).
Let's compare these to Eliud's 2:01:39 best.
2:01:39 / 2:13:13 = 7299s / 7993s = .9132
2:01:39 / 2:17:23 = 7299s / 8243s = .8855
If we assume that these ratios hold more or less constant as we extend from the marathon distance to 100k, what would Eliud's performance in the 100k be? (Of course, this is a big assumption. It depends on Kazami's marathon bests being reflective of his top fitness, and even beyond that one could argue either way, that the ratio should actually decrease as we elongate the distance, or that it should increase. Hard to say precisely, but we're just looking for a range of possibilities here).
We have .9132 * 6:09:14 = .9132 * 22,154s = 20,231s = 5:37:11
and .8855 * 6:09:14 = .8855 * 22,154s = 19,617s = 5:26:27
This predicts a plausible range of 5:26 to 5:37 for Eliud in the 100k. Now, of course, this kind of projection could be meaningless - if I compare two runners in the 1500m or 5000m and try to extrapolate up to the marathon, we could in fact see an inversion rather than a stability in their performance ratio, as runners have different aptitudes at different distances. I would tend to be more confident in extrapolating from 42k to 100k than in 1500 to 42k because of the similar physiological profiles of the races, but you could easily argue against this: the 100k certainly presents unique demands beyond just being a "longer marathon".
Still, I like to think that these numbers present a plausible ability range for Eliud. If you believe he has substantially superior endurance (from 42k to 100k) than Kazami, you could imagine him running as fast as 5:15 or so. If you believe he has substantially inferior endurance, perhaps he runs as slow as 5:45, 5:50. We are unfortunately flying blind here, due to the almost complete dearth of top tier marathon talent tackling the ultra events in their prime.
For what its worth, 5:15 is 5:04 pace, 5:26 is 5:15 pace, 5:37 is 5:25 pace, and 5:50 takes you to 5:37 pace. A human being running for 5 and a half hours at 5:20 pace does not seem possible, but its hard to overstate how much better 2:01:39 is than 2:13 or 2:17.
If we scale this performance differential of roughly 90% down to the mile distance, you are talking about the difference between say a 3:54 runner and a 4:20 miler. Meditate for a moment on the performance discrepancy in ceilings you would expect in the marathon for a 3:54 versus a 4:20 runner. That is what we are dealing with in the case of Kipchoge versus Kazami. Assuming both runners are distance runners (i.e., the 3:54 guy is not a 400/800m specialist or whatever), there is really no conceivable reality in which you would expect the 4:20 guy to beat the 3:54 guy over the marathon distance. The difference in ceilings is simply too vast.